BUG: Let MultiIndex.set_levels accept any iterable (#23273) #23291
Conversation
|
Hello @imankulov! Thanks for submitting the PR.
|
Codecov Report
@@ Coverage Diff @@
## master #23291 +/- ##
==========================================
+ Coverage 92.22% 92.22% +<.01%
==========================================
Files 169 169
Lines 50911 50913 +2
==========================================
+ Hits 46954 46956 +2
Misses 3957 3957
Continue to review full report at Codecov.
|
jreback
left a comment
jreback
left a comment
There was a problem hiding this comment.
pls add a whatsnew note in bug fixes / indexing
| levels = list(levels) | ||
|
|
||
| if level is not None and not is_list_like(level): | ||
| if not is_list_like(levels): |
There was a problem hiding this comment.
you can just add
elif is_list_like(levels):
levels = list(levels)
There was a problem hiding this comment.
Tried this approach, but unfortunately, it breaks functionality where one passes to set_levels CategoricalIndex (or in general, something else which is more complex than a list).
Specifically, this turns
index.set_levels(CategoricalIndex(list("bac")), 0)into
index.set_levels(list("bac"), 0)and breaks test_set_levels_categorical
I can add change this with
if is_list_like(levels) and not isinstance(levels, Index):but I'm not sure what's better
There was a problem hiding this comment.
the 2nd would be fine, we don't need/want any more api checkers)
| new_sizes = map(int, ['3', '2', '1']) | ||
| index = pd.MultiIndex.from_arrays([sizes, colors], names=['size', 'color']) | ||
| index2 = index.set_levels(new_sizes, level='size') | ||
| assert_matching(index2.get_level_values('size'), [3, 2, 1]) |
There was a problem hiding this comment.
construct the appropriate MultiIndex and use tm.assert_index_equal
pandas/core/dtypes/inference.py
Outdated
| Check if the object is subscriptable. | ||
| String types are not included as sequences here. | ||
| Parameters |
|
Thanks for the review @jreback. Addressed all comments. Need advice on how to deal better with set_index(CategoricalIndex) |
| levels = list(levels) | ||
|
|
||
| if level is not None and not is_list_like(level): | ||
| if not is_list_like(levels): |
There was a problem hiding this comment.
the 2nd would be fine, we don't need/want any more api checkers)
| index = pd.MultiIndex.from_arrays([sizes, colors], names=['size', 'color']) | ||
|
|
||
| new_sizes = map(int, ['3', '2', '1']) | ||
| index = index.set_levels(new_sizes, level='size') |
|
thanks! |
3 participants
git diff upstream/master -u -- "*.py" | flake8 --diffQuite surprisingly,
is_list_likeaccepts a wide range of iterables including non-subscriptable objects. Later in the codeset_levelsimplicitly supposes that passed instance can return its 0'th element by index, which is not always the case, as provided in the example of the #23273.To address the issue, a new helper function
is_subscriptableis added, and if passed levels look like a list, but are not subscriptable, they are explicltly converted to a list withlist(levels)